Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Content-Type: text/plain;
        charset=utf-8
Mime-Version: 1.0 (1.0)
Subject: Re: [PATCH v3] x86/vdso: Handle clock_gettime(CLOCK_TAI) in vDSO
From:   Andy Lutomirski <luto@amacapital.net>
In-Reply-To: <5f9e890c-fd67-2e39-42ca-64290301b20a@redhat.com>
Date:   Thu, 13 Sep 2018 12:35:54 -0700
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Andy Lutomirski <luto@kernel.org>,
        Matt Rickard <matt@softrans.com.au>,
        Stephen Boyd <sboyd@kernel.org>,
        John Stultz <john.stultz@linaro.org>, X86 ML <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <B47AF238-71A6-4FC1-BC5A-55893C2D76B0@amacapital.net>
References: <20180901015935.CCF0B18E20A9@virtux64.softrans.com.au> <CALCETrU4PzskO9pE6G4EkXNRRX3Mu7Vvjv1P8XgrfRMdgGDZSg@mail.gmail.com> <alpine.DEB.2.21.1809092159440.1402@nanos.tec.linutronix.de> <dde36e93-a21b-6223-8ffc-6dd11f687d38@redhat.com> <alpine.DEB.2.21.1809121616040.9957@nanos.tec.linutronix.de> <59403142-1073-3926-65dd-ac55ead357b9@redhat.com> <alpine.DEB.2.21.1809121623430.9957@nanos.tec.linutronix.de> <DEC1E2DC-2D89-4955-81FC-4EDCF3692214@amacapital.net> <0120ea99-fcac-839a-ef8f-1daa74c17bec@redhat.com> <F5CE2714-B491-490D-9B03-E111A61E333F@amacapital.net> <5f9e890c-fd67-2e39-42ca-64290301b20a@redhat.com>
To:     Florian Weimer <fweimer@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


> On Sep 13, 2018, at 12:07 PM, Florian Weimer <fweimer@redhat.com> wrote:
>=20
> On 09/13/2018 05:22 PM, Andy Lutomirski wrote:
>>> On Sep 13, 2018, at 1:07 AM, Florian Weimer <fweimer@redhat.com> wrote:
>>>=20
>>> On 09/12/2018 07:11 PM, Andy Lutomirski wrote:
>>>>> The multiplexer interfaces need much more surgery and talking about fu=
tex,
>>>>> we'd need to sit down with quite some people and identify the things t=
hey
>>>>> actually care about before just splitting it up and keeping the existi=
ng
>>>>> overloaded trainwreck the same.
>>>>>=20
>>>> There=E2=80=99s also the issue of how much the speedup matters. For fut=
ex, maybe a better interface saves 3ns, but a futex syscall is hundreds of n=
s. clock_gettime() is called at high frequency and can be ~25ns. Saving a fe=
w ns is a bigger deal.
>>>=20
>>> My concern is that the userspace system call wrappers currently do not k=
now how many arguments the individual operations take and what types the arg=
uments have (hence the =E2=80=9Ctype-polymorphic=E2=80=9D nature I mentioned=
). This could be a problem for on-stack argument passing (where you might re=
ad values beyond the end of the stack, and glibc avoids that most of the tim=
e by having enough cruft on the stack), and for architectures which pass poi=
nters and integers in different registers (like some m68k ABIs do for the re=
turn value).
>=20
>> Isn=E2=80=99t clock_gettime already special because of the vDSO entry poi=
nt, though?
>=20
> Somewhat special, yes, but not overly so, and not in the type-polymorphic s=
ense.  We can't give direct access of the vDSO implementation to application=
s because the kernel does not know about the userspace errno variable.  We d=
o that for time on x86_64, where applications call into the vDSO directly, b=
ypassing glibc completely after binding.

If the vDSO adds special helpers for CLOCK_MONOTONIC and CLOCK_REALTIME, I t=
hink we can reasonably safely promise that they never fail. (seccomp can obv=
iously break that promise if there=E2=80=99s no TSC, but I think that seccom=
p users who do that get to keep both pieces.)